To illustrate message passing consider the case of determining the path to send a message from processor 0 to processor 7 in a 3-dimensional hypercube as shown in Figure 2.19. If the message is to traverse a path which is of minimal length, that is d (0, 7), then it should travel over three edges. For this case there are six possible paths: Figure 2.19 Hypercube Architecture In general, in a hypercube of dimension d, a message travelling from processor x to processor y has d (x, y) ! distinct paths (see Problem 2.11). One simple algorithm is to compute the exclusive-or of the source and destination processors and traverse the edge corresponding to complementing the first bit that is set. This is illustrated in Table 2.4 for left to right complementing and in Table 2.5 for right to left complementing. Table2.4Calculating theMessagePath— LefttoRightProcessor Source ProcessorDestination Exclusive‐Or NextProcessor 000 111 111 100 100 111 011 110 110 111 001111 Table2.5CalculatingtheMessagePath—Right toLeftProcessorSource Processor Destination Exclusive‐ Or Next Processor 000 111 111 001 001 111 110 011 011 111 100111 The message passing algorithm still works under certain circumstances even when the hypercube has nodes that are faulty. This is discussed in the next section. 2.6.3EfficientHypercubes This section presents the analysis of the class of hypercubes for which the message passing routines of the previous section are valid. Examples are presented in detail for an 8-node hypercube. 2.6.3.1TransitiveClosure Definition 2.23 The adjacency matrix, A, of a graph, G, is the matrix with elements a ij such that a ij = 1 implies there is an edge from i to j. If there is no edge then a ij = 0. The adjacency matrix, A, of the transitive closure of the 8-node hypercube is simply the matrix For a hypercube with all functional nodes every processor is reachable. Previous TableofContents Next Copyright © CRC Press LLC Algorithms and Data Structures in C++ by Alan Parker CRC Press, CRC Press LLC ISBN: 0849371716 Pub Date: 08/01/93 Previous TableofContents Next 2.6.3.2LeastWeightedPathLength Definition 2.24 The least-weighted path-length graph is the directed graph where the weights of each edge correspond to the shortest path-length between the nodes. The associated weighted matrix consists of the path-length between the nodes. The path-length between a processor and itself is defined to be zero. The associated weighted matrix for an 8- node hypercube with all functional nodes is a ij is the distance between nodes i and j. If nodes i and j are not connected via any path then a ij = ∞. 2.6.3.3HypercubeswithFailedNodes This section introduces the scenario of failed processors. It is assumed if a processors or node fails then all edges incident on the processor are removed from the graph. The remaining processors will attempt to function as a working subset while still using the message passing algorithms of the previous sections. This will lead to a characterization of subcubes of a hypercube which support message passing. Consider the scenario illustrated in Figure 2.20. In the figure there are three scenarios with failed processors. In Figure 2.20b a single processor has failed. The remaining processors can communicate with each other using a simple modification of the algorithm which traverses the first existing edge encountered. Similarly, in Figure 2.20c communication is still supported via the modified algorithm. This is illustrated in Table 2.6. Notice that in Table 2.6 the next processor after 000 was 001. For the topology in the figure the processor did not exist so the algorithm proceeded to the next bit from right to left which gave 010. Since this processor existed the message was sent along the path. Figure 2.20 Hypercube with Failed Nodes Table2.6Calculating theMessagePath— RighttoLeftforFigure 2.20cProcessor ProcessorDestination Exclusive‐Or NextProcessor Source 000 111 111 010 010 111 101 011 011 111 100111 The scenario in Figure 2.20d is quite different. This is illustrated in Table 2.7. In this case, the first processor considered to is 001 but it is not functional. Processor 010 is considered next but it is not functional. For this case the modified algorithm has failed to route the message from processor 000 to 011. There exists a path from 000 to 011 one of which is Notice that the distance between the processors has increased as a result of the two processors failures. This attribute is the motivation for the characterization of efficient hypercubes in the next section. Table2.7Calculating theMessagePath— RighttoLeftforFigure 2.20dProcessorSource ProcessorDestination Exclusive‐Or NextProcessor 000 011 011? 2.6.3.4Efficiency Definition 2.25 A subcube of a hypercube is efficient if the distance between any two functional processors in the subcube is the same as the distance in the hypercube. A subcube with this property is referred to as an efficient hypercube. This is equivalent to saying that if A represents the least-weighted path-length matrix of the hypercube and B represents the least-weighted path-length matrix of the efficient subcube then if i and j are functional processors in the subcube then b ij = a ij . This elegant result is proven in Problem 2.20. The least-weighted path-length matrix for efficient hypercubes place ∞ in column i and row i if processor i is failed. The cubes in Figure 2.20b and c are efficient while the cube in Figure 2.20d is not efficient. If the cube is efficient then the modified message passing algorithm in the previous section works. The next section implements the procedure for hypercubes with failed nodes. 2.6.3.5MessagePassinginEfficientHypercubes The code to simulate message passing in an efficient hypercube is shown in Code List 2.8. The output of the program is shown in Code List 2.9. The path for communicating from 0 to 63 is given as 0-1-3-7-15-31-63 as shown in Code List 2.9. Subsequently processor 31 is deactivated and a new path is calculated as 0-1-3-7-15-47-63 which avoids processor 31 and traverses remaining edges in the cube. The program continues to remove nodes from the cube and still calculates the path. All the subcubes created result in an efficient subcube. Code List 2.8 Message Passing in an Efficient Hypercube Code List 2.9 Output of Program in Code List 2.8 2.6.4VisualizingtheHypercube:AC++Example This section presents a C++ program to visualize the hypercube. A program to visualize the cube is shown in Code List 2.10. The program was used to generate the PostScript image in Figure 2.21 for a 64 node hypercube. The program uses a class structure similar to the program to visualize the Tower of Hanoi in Code List 2.6. The program introduces a new PostScript construct to draw and fill a circle x y radius angle1 angle2 arc The program uses the scale operator to force the image to fill a specified area. To illustrate this, notice that the program generated both Figure 2.21 and Figure 2.22. The nodes in Figure 2.22 are enlarged via the scale operator while the nodes in Figure 2.21 are reduced accordingly. The strategy in drawing the hypercube is such that only at most two processors appear in any fixed horizontal or vertical line. The cube is grown by replications to the right and downward. Figure 2.21 A 64-Node Hypercube Code List 2.10 C++ Code to Visualize the Hypercube Figure 2.22 An 8-Node Hypercube Code List 2.11 Output of Program in Code List 2.10 Previous TableofContents Next Copyright © CRC Press LLC Algorithms and Data Structures in C++ by Alan Parker CRC Press, CRC Press LLC ISBN: 0849371716 Pub Date: 08/01/93 Previous Table of Contents Next 2.7 Problems (2.1) [Infinite Descent — Difficult] Prove, using infinite descent, that there are no solutions in the positive integers to (2.2) [Recuffence] Find the closed form solution to the recursion relation and write a C++ program to calculate the series via the closed form solution and print out the first twenty terms of the series for (2.3) [Tower of Hanoi] Write a C++ Program to solve the Tower of Hanoi problem for arbitrary n. This program should output the move sequence for a specific solution. (2.4) [Tower of Hanoi] Is the minimal solution to the Tower of Hanoi problem unique? Prove or disprove your answer. (2.5) [Rectangular Mesh] Given an 8x8 rectangular mesh with no additional edge connections calculate the largest distance between two processors, where the distance is defined as the minimum number of edges to traverse in a path connecting the two processors. (2.6) [Rectangular Mesh] For a rectangular mesh with no additional edge connections formally describe the topology in terms of vertices and edges. (2.7) [Rectangular Mesh] Write a C++ program to generate a PostScript image file of the rectangular mesh for 1 ≤ n ≤ 20 without additional external edge connections. To draw a line from the current point to (x, y) use the primitive followed by to actually draw the line. Test the output by sending the output to a PostScript printer. (2.8) [Cube-Connected Cycles] Calculate the number of edges in a cube connected cycles topology with nlog n nodes. (2.9) [Tree Structure] For a graph G, which is a tree, prove that (2.10) [Cube-Connected Cycles] For a cube-connected cycles topology formally describe the topology in terms of vertices and edges. (2.11) [Hypercube] Given two arbitrary nodes in a hypercube of dimension n calculate the number of distinct shortest paths which connect two distinct nodes, A and B, as a function of the two nodes. Use a binary representation for each of the nodes: (2.12) [Hypercube] Given a hypercube graph of dimension n and two processors A and B what is the minimum number of edges that can be removed such that there is no path from A to B. (2.13) Is every edge in a tree a bridge? (2.14) Devise a broadcast algorithm for a hypercube of arbitrary dimension. Write a C++ program to simulate this broadcast operation on an 8-dimensional hypercube. (2.15) Devise a message passing algorithm for a hypercube of arbitrary dimension. Write a C++ program to simulate this algorithm and demonstrate it for a 12-dimensional hypercube. (2.16) Write a C++ program to visualize a complete binary tree. Your program should scale the node sizes to fit on the page as a function of the dimension in a similar fashion to Code List 2.10. (2.17) Describe in detail the function of each procedure in the code to visualize the hypercube in Code List 2.10. Present a high-level description of the procedures render_cube and init_cube. (2.18) Write a C++ program to display the modified adjacency matrix of an n- dimensional hypercube similar to the matrix presented in Eq. 2.67. (2.19) Write a C++ program to visualize a 64-node hypercube which supports message passing. Your program should use a separate gray level to draw the source and destination processors and should draw the edges which form the path in a different gray scale also. (2.20) [Difficult] Prove that the modified message passing algorithm works for any two functional processors in an efficient hypercube. (2.21) Write a C++ program to determine if a hypercube with failed nodes is efficient. (2.22) Calculate the least-weighted path-length matrix for each of the subcubes in Figure 2.20. (2.23) Given a hypercube of dimension d calculate the probability that a subcube is efficient where the subcube is formed by the random failure of two processors. (2.24) Modify the C++ program in Code List 2.10 to change the line width relative to the node size. Test out the program for small and high dimensions. (2.25) Rewrite Code List 2.10 to build the hypercube using a recursive function. (2.26) The program in Code List 2.10 uses a simple algorithm to draw a line from each processor node to its neighbors. As a result, the edges are drawn multiple times within in the file. Rewrite the program to draw each line only once. Previous Table of Contents Next Copyright © CRC Press LLC . 110 110 111 001 111 Table2.5CalculatingtheMessagePath—Right toLeftProcessorSource Processor Destination Exclusive‐ Or Next Processor 000 111 111 001 001 111 110 . left complementing. Table2.4Calculating theMessagePath— LefttoRightProcessor Source ProcessorDestination Exclusive‐Or NextProcessor 000 111 111 100 100 111 011 110 110 . by to actually draw the line. Test the output by sending the output to a PostScript printer. (2.8) [Cube-Connected Cycles] Calculate the number of edges in a cube connected cycles topology